A correspondent pointed me to this document, dated March 30, 1965, in which an executive with Western Union, the telegraph company, lays out the company’s ambitious plan to create “a nationwide information utility, which will enable subscribers to obtain, economically, efficiently, immediately, the required information flow to facilitate the conduct of business and other affairs.”
The idea of a “computing utility” was much discussed in the 1960s, but this document nonetheless provides a remarkably prescient outline of what we now call cloud computing. Some excerpts:
Over the past century or more there have evolved in this country a limited number of basic systems serving the general public – a group generally termed “public utilities.” These utilities serve, among others, such fields as transportation; communications (telegraph, telephone, cable, radio, the broadcast services, etc.); and the energy systems, distributing power.
What is now developing, very rapidly, is a critical need – as yet not fully perceived – for a new national information utility which can gather, store, process, program, retrieve and distribute on the broadest possible scale, to industry; to the press; to military and civilian government; to the professions; to department stores, banks, transportation companies and retailers; to educational institutions, hospitals and other organizations in the fields of public health, welfare and safety; and to the general public, virtually all of the collected useful intelligence available, through locally-, regionally- and nationally-linked systems of computers. Just as an electrical energy system distributes power, this new information utility will enable subscribers to obtain, economically, efficiently, and immediately, the required information flow to facilitate the conduct of business, personal and other affairs.
There is no substantial technical bar even now to the establishment of such a nationwide information utility. Computers and associated equipment, the methodology, the storage and retrieval techniques, the knowledge required to provide the very broad bandwidth required for high-speed data transmission – all these exist today. Their harnessing into a national system presents no technical problems essentially more difficult than the strategic placements a half-century ago of steam turbines to create electrical energy, and the related building of power grids … Indeed, the computer and the turbine share a common characteristic in that (within appropriate limits of optimum sizes and capacities) the larger the unit, the more efficient it is in terms of unit-cost production … The cardinal economic principle at issue here is that an information utility serving a large number of users can provide service to each more economically than he can provide it for himself, just as a power system can provide energy to its customers at lower cost than they, individually, can generate it for themselves …
We envision, then, the expansion of the existing plant, offices, personnel, and nationwide operations of Western Union, to transform it into a national information system [that] would furnish a uniform, efficient, integrated information service to meet the needs of all types of users, everywhere …
It might be added, here, that any movement by the Bell System to substitute itself for Western Union as the nation’s information utility, as well as the pervasive, dominant power in the telephone field, would obviously create profound national concern on the score of “giantism” – since any further and large assumption of added power would bring about one entity of even more menacing size than now …
Western Union has the skills and experience that uniquely qualify if for such a role; the public need for such a new utility is growing at a rapid rate; the field is already large and the potential tremendous – probably at least as large as any other national utility that exists today.
When the history of cloud computing is written, it may be that Western Union will play the role that Xerox now plays in the history of the personal computer: the company that saw the future first, but couldn’t capitalize on its vision.
There was a large number of companies in the late 60s and early 70s offering cloud computing. It was called Timesharing. Same basic idea, large scale computers professionally managed and operated, accessed via cheap devices over telecommunications.
ADP, Compuserv, IBM, and others were in the business, including one name First Data Corp, which is in a completely unrelated business today.
If you read the document, what Western Union are describing is fundamentally different from the timeshare systems of that era – they talk about broadband networks, computing clusters, et. al.
Yes, while time-sharing was clearly an inspiration for the concept of utility computing, what Western Union is laying out here is very different and much closer to today’s conception of cloud computing.
Here, by the way, is another document from the sixties that will be of interest to the historically minded.
And don’t forget Parkhill’s Challenge of the Computer Utility, published in 1966.
You have to wonder if the Internet and the WWW had caught on before the PC did, whether the trajectory of the last 20 years of IT would have been very different.
It may be that in 20 years, we’ll look back and decide that the whole desktop computer / packaged software era was the big anomaly.
As our Fearless Leader has pointed out, cloud computing only became practial when the network speeds approached that of the local computer bus. For those of us whos first taste of the Internet was dialuup in the 1990s, this video show how little the technology had changed between 1964 and the mid-1990s(thriet years!):
1964 Antique MODEM Live Demo. The change from low bandwidth over the PSTN to braodband was similar to the change from DC to AC during the early development of electrical grid power.
Mr. Carr,
If you can, please try to interview Kernighan, Thompson, Ritchie, et al. about the utility computing ideas that *were* tossed around at Bell by at least the ’70s. I’m frustrated that I can’t quite quickly find source materials or first-hand accounts to point you at but I’m certain I’ve read:
There was a notion (didn’t get *too* far) at Bell of “Personal Computing” by which they meant building clusters of timesharing computers at central offices and putting smart terminals (aka thin clients) in every home – a bit like the French minitel system but much fancier. They had the basic notion of big “plants” that manufactured cycles and storage, accessed over a distribution network that would happen to coincide with Bell’s existing phone network. An example of late motion in that direction was the Blit terminal. Another example of motion in that direction was the (by then busted up) AT&T initial foreys into data-over-voice networks (initially, always-on fast modems to terminal switches, later turning into DSL). But the idea goes back to fairly early on in the history of unix, as I recall. May have even related to some of the motivation behind Multics.
-t
I don’t see it as a fundamental difference. “Western Union is laying out here is very different and much closer to today’s conception of cloud computing”. What they describe is exactly what the major players in the 70s delivered. CompuServ’s service for PCs was simply a way to use the left-over cycles and networking after their business customers went home.
Perhaps I don’t see it because what a lot of today’s cloud computing hype is nothing more than a 40 year later spin on what good timesharing provided. Its not much of a jump looking at the the limited abilities of a netbook or Google Chrome computer talking to a big network and how that parallels the smarter terminals of the 70s and timesharing. I expect most of today’s tech writers never got exposure to what was being done with Multics or Tenex in the 70s, or even the French Minitel in the 80s.
Not as cool, no live video, but Minitel delivered a lot of the things promised decades later in the dot.com boom.
There is a lot of talk about the 40th anniversary of the invention of the Arpanet. Almost all of that work was done on Multics and Tenex systems.